Gen AI Learning Resources
This page provides an overview of the basic skills one needs to use generative AI tools safely and effectively. It presupposes a basic awareness of several topics foundational to generative AI, including tokens, context, and retrieval-augmented generation (RAG). For an introduction to these and other foundational topics, watch Generative AI Basics.
Do not use your BC credentials (BC username, password, or any BC email address) to sign up for non-BC-approved generative AI tools. For a list of approved tools and instructions on how to access them, visit Gen AI Tools at BC Law.
Prompt and Context Engineering
You’ve probably heard of prompt engineering: writing clear instructions for an AI model to get the desired output. But a prompt is only one part of what a model sees when it generates an output. The full picture is context—which is why practitioners increasingly talk about context engineering instead.
Good prompting is still important. But getting more from these tools—whether you’re analyzing documents or building an automated workflow—involves managing the rest of the context too. This can include files you upload, earlier messages in the conversation, web search results, saved memories, and more. Context engineering requires an understanding of what the model can see in any given session, as well as what the task does and does not require.
Learn more
- Prompting Fundamentals
- Elements of a Prompt
- Artificial Intelligence in Legal Research: Effective Prompting Tips
- Context Engineering Fundamentals
- Context Engineering Guide
- Prompt Engineering: How to Talk to the AIs (LinkedIn Learning)
BC provides access to LinkedIn Learning; first-time users should sign in to BC LinkedIn Learning.
Choosing the Right Model for the Task
Most generative AI platforms allow users to choose among several models when starting a work session.
Different models excel at different tasks and consume different amounts of tokens—the units of meaning that generative AI models process and the basis on which companies set pricing and usage limits. Understanding which model suits which task is therefore essential to getting the best result for the fewest tokens (lowest cost).
Many free online resources exist that provide advice on this topic; a few are included below.
- LLM Leaderboard - Comparison of AI Models
- Chat GPT vs Claude vs Gemini
- How to Choose the Right AI Model for Your Use Case
Do not use your BC credentials (BC username, password, or any BC email address) to sign up for non-BC-approved generative AI tools. For a list of approved tools and instructions on how to access them, visit Gen AI Tools at BC Law.
Learn More
- Reasoning models vs standard LLMs: Which one should handle each request?
- General AI vs Legal AI: Key Differences
Validating Outputs & Catching Hallucinations
Generative AI models are designed to produce fluent, plausible-sounding text, not to verify truth. Even the most capable models occasionally fabricate facts, misconstrue source documents, or invent citations, and they often do so with complete confidence. Treating every AI output as a first draft rather than a finished product and verifying all factual claims are essential to using these tools responsibly.
The level of scrutiny an output requires depends on the nature and stakes of the task, but several practical steps can help you proactively catch errors:
Whenever possible, upload your own documents or paste the relevant text into the conversation and instruct the model to rely exclusively on that material. In addition, instruct it to quote or cite specific passages for each claim so that you can quickly verify its accuracy.
- Never assume an AI output (e.g., a case citation, statute, academic reference, URL, line of code, or calculation) is accurate or that it meets accepted standards. Whenever accuracy or quality matters to the task at hand, verify the output independently.
- Apply lateral reading, the practice of searching for and using independent and trusted sources to corroborate or debunk the model’s output.
- Defend against model sycophancy, the tendency of AI models to tell users what they want to hear rather than prioritize accuracy. You can mitigate, though not eliminate, this tendency by instructing a model to critique your work, play devil’s advocate, or offer alternative perspectives.
Learn More
- Artificial Intelligence (AI): Evaluating AI Sources
- Seeing real value from AI depends on being able to verify its outputs
Using Agent “Skills”
Most major consumer AI platforms now support agent skills, reusable packages of instructions, reference files, and scripts (or some combination thereof) that “teach” an AI agent to perform a specific task or workflow. If you find yourself instructing AI models to do the same task over and over again, you might consider creating a skill for that task to prevent the need to continue repeating yourself and free up time for other work.
Ideas for agent skills:
- University Brand and Style Editor: Edits drafts to match an institution’s official editorial style guide.
- Digital Accessibility (ADA) Checker: Reviews documents, emails, and web copy to flag accessibility issues.
- Meeting Minutes and Action Item Extractor: Converts messy meeting transcripts or rough notes into a clean summary organized by key decisions, owners, and deadlines.
Most of these platforms now include built-in “skill creator” skills that let you build reusable skills simply by describing what you want in plain language. Most also launch skills automatically whenever your prompt matches a skill’s description. Skills can be launched manually too.
Agent skills are a vendor-neutral, open standard, which means the core instructions of a skill you build in one AI platform can often be reused in others as well.
Learn More
- Agent Skills: What are they and how do they work? (Basic)
- Introduction to Agent Skills (Basic)
- Building skills for AI agents: pitfalls and best practices for skill development (Intermediate)
Further Learning Resources
Artificial Intelligence in Legal Research
- An introduction to generative AI research tools and how they're used in law school and in practice.
Copyright, Authors' Rights, and Generative AI
- An overview of the deals that academic publishers are making with AI developers, as well as guidance for scholars on how to navigate the changing landscape of academic publishing.
A self-paced, online program created by teaching and research librarians to help BC Law students master core legal research skills, including the effective and ethical use of generative AI, for practice and the NextGen Bar Exam. Demonstrate your expertise by adding a verified credential to your LinkedIn profile.
Short video courses on practical AI topics, from LLM fundamentals and prompting to risks and legal ethics. BC Law students can follow a customized track to a Generative AI Fundamentals Certificate.
A customized, self-paced learning track for BC Law students who want a general introduction to generative AI: what it is, its legal applications, its limitations and risks, and more.
AI @ BC has a wealth of resources on generative AI, including:
- LinkedIn Learning Trainings — free for BC students, faculty, and staff. See Build Your Skills with LinkedIn Learning to get started.
- Engaging with AI — resources on generative AI as it relates to teaching and learning, developed by the Center for Digital Innovation in Learning.
- Artificial Intelligence in Teaching and Learning — a guide from the Center for Teaching Excellence.
- Guidance for Appropriate Use of AI in Research — from the Office of the Vice Provost for Research.
See AI @ BC for the full list.
